As part of my ongoing Prosper investment endeavors, I thought I'd post some research I performed this weekend. In trying to come up with an optimal lending strategy, I decided to look at what impact the amount of money borrowed would have on default rates. My thinking was that all other things being equal, an individual would be less prone to default on a loan with a smaller amount as opposed to a larger amount.

Here's why:

• It's easier to pay back small a loan, and defaulting on it is less attractive due to credit impact.


• It's more tempting to take the money and run or default on say a 25K loan compared to a 2K loan. For a 2K loan, it wouldn't be worth the hit to your credit, whereas someone that already has trouble managing their money may not think twice about pocketing a quick 25K without regard for their credit score.


• In the event there's some sort of fraud, impersonation, scam, etc. that sort of elaborate ruse would certainly be worth it for a higher amount (you don't see people counterfeiting $1 bills; they're always $20 and up).

I used the Performance tool on the Prosper site and due to the wide date range, the data sets should contain several hundred loans each, rendering the sample size to be adequate to draw meaningful conclusions. One disclaimer: True default rates will likely exceed those listed here since the loans are all less than 3 years old. Over time, loans default...even if they're paying on time initially. So, a 1 year loan data set might show a 1% default rate, but over the next two years, it's likely that more loans would default since it is a reality that some people lose the ability or desire to pay over time as conditions in their lives change. The intent of this exercise was to study the trends and what these mean to contribute to successful lending. It is not meant to be a firm endorsement for actual projected default rates, but rather see how the trends go.

Any analysis requires some boundary conditions and assumptions. Here were the ones I used:

• Jun 12, 2005- Feb 12 2007 to allow time for recent loans to default


• No current delinquencies


• Unlimited recent inquiries


• No other criteria (includes groups and non groups, etc)

I broke out the groups by low loan amounts (up to 3000) and medium and high, etc. as indicated on chart.

Based on these criteria, it was difficult to get a lot of Ds and Es, so the numbers drop out at the higher levels.


Analysis of Results:

• At the AA and A credit grades, there appears to be a pretty firm trend that people with lower loan amounts have a much lower tendency to default, supporting my hypothesis (which is actually conventional wisdom in the lending industry; it's just that Prosper's a bit of a new entity, so conventional wisdom doesn't always hold true - hence the higher default rates than industry).



• The low loan amount trend seems to hold for all credit grades with some minor overlaps likely due to statistical sample error.



• The 10-25K loans consistently exceed default rates across all credit grades and are unacceptably high compared to lower amounts from C to worse.



• In the low loan amount class of 3000 and less, it's interesting to note that B and C loans have about the same default rate. Given the fact that you can attain higher interest rates on the Cs, may want to preferentially target Cs over Bs for these low loan amounts.



• These trends aren't entirely surprising, it's just that there's now been some light shed on some minor trends which broke out from the expected norms. Markets are efficient, but this one isn't yet due to the human nature aspect (personalized details, multiple variables in each loan profile). Therefore analyses like this can help you exploit inefficiencies in the risk/return profiles of various lending grades/amounts.

$25 Signup bonus for new members for a few more weeks:

Reminder: If you haven't done so already, there are only a few weeks left to sign up and get the $25 referral fee as a new member lender by clicking on the link on the left. I too will then get $25. Once you're a member, you can forward email links or add banners to your blogs to participate in the referral awards through August. If borrowing, click on the link at the bottom. Thanks for supporting the site; to date, averaging about a referral per week. Seeing at least some compensation makes performing this research, formatting and posting worthwhile.

And don't forget to take the Prosper Poll on the left by July 22!
.